#!git clone https://github.com/TPM034A/Q2_2025
#!pip install -r Q2_2025/requirements_colab.txt
#!mv "/content/Q2_2025/Lab_sessions/lab_session_01/data" /content/data

# Import required Python packages and modules
import os
import numpy as np
import pandas as pd
import geopandas as gpd
import seaborn as sns
import matplotlib.pyplot as plt


# Import selected functions and classes from Python packages
from os import getcwd
from pathlib import Path
from sklearn.linear_model import LinearRegression
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error, r2_score

# Get the current working directory
working_folder = getcwd()
data_folder = Path(f'data')
print(data_folder)

data

# Load liveability data, using pandas
gemeente_liv_data = pd.read_csv(data_folder/'gemeente_liveability.csv')
wijk_liv_data = pd.read_csv(data_folder/'wijk_liveability.csv')
buurt_liv_data = pd.read_csv(data_folder/'buurt_liveability.csv')
print(list(gemeente_liv_data.keys()))

['gm_code', 'versie', 'jaar', 'gm_naam', 'lbm', 'afw', 'fys', 'onv', 'soc', 'vrz', 'won']

# Load files on population statistics
gemeente_pop_data = pd.read_csv(data_folder/'gemeente_2020_pop.csv', sep=';')
wijk_pop_data = pd.read_csv(data_folder/'wijk_2020_pop.csv', sep=';')
buurt_pop_data = pd.read_csv(data_folder/'buurt_2020_pop.csv', sep=';')
gemeente_pop_data.keys()

Index(['GM_CODE', 'GM_NAAM', 'H2O', 'OAD', 'STED', 'BEV_DICHTH', 'AANT_INW',
       'AANT_MAN', 'AANT_VROUW', 'P_00_14_JR',
       ...
       'AF_PODIUM', 'AV5_PODIUM', 'AV10PODIUM', 'AV20PODIUM', 'AF_MUSEUM',
       'AV5_MUSEUM', 'AV10MUSEUM', 'AV20MUSEUM', 'JRSTATCODE', 'JAAR'],
      dtype='object', length=178)

# Load shape files
gemeente_shape = gpd.read_file(data_folder/'gemeente/gemeente 2020.shp')
wijk_shape = gpd.read_file(data_folder/'wijk/wijk 2020.shp')
buurt_shape = gpd.read_file(data_folder/'buurt/buurt 2020.shp')
print(buurt_shape.info())

<class 'geopandas.geodataframe.GeoDataFrame'>
RangeIndex: 13808 entries, 0 to 13807
Data columns (total 4 columns):
 #   Column    Non-Null Count  Dtype   
---  ------    --------------  -----   
 0   bu_code   13808 non-null  object  
 1   bu_naam   13808 non-null  object  
 2   gm_naam   13808 non-null  object  
 3   geometry  13808 non-null  geometry
dtypes: geometry(1), object(3)
memory usage: 431.6+ KB
None

# Ensure that the shape files are all in the right projection
gemeente_shape.to_crs(epsg=28992, inplace=True)
wijk_shape.to_crs(epsg=28992, inplace=True)
buurt_shape.to_crs(epsg=28992, inplace=True)

# Print crs info
print(gemeente_shape.crs)
print(wijk_shape.crs)
print(buurt_shape.crs)

EPSG:28992
EPSG:28992
EPSG:28992

# Remove rows where lbm is NaN
gemeente_liv_data.dropna(subset = ['lbm'], inplace = True)

# Extract subsets: only year 2020 and Leefbaarometer 3.0 data
gemeente_liv_data = gemeente_liv_data.loc[(gemeente_liv_data.versie ==  'Leefbaarometer 3.0') & (gemeente_liv_data.jaar == 2020)]
wijk_liv_data.dropna(subset = ['lbm'], inplace = True)
wijk_liv_data = wijk_liv_data.loc[(wijk_liv_data.versie ==  'Leefbaarometer 3.0') & (wijk_liv_data.jaar == 2020)]
buurt_liv_data.dropna(subset = ['lbm'], inplace = True)
buurt_liv_data = buurt_liv_data.loc[(buurt_liv_data.versie ==  'Leefbaarometer 3.0') & (buurt_liv_data.jaar == 2020)]

def clean_pop_data(pop_data):

    # Extract subsets: only data where is land
    pop_data  = pop_data.loc[pop_data.H2O == 'NEE'] 

    # Remove rows where P_KOOPWON == 0 or P_HUURWON == 0
    pop_data = pop_data.loc[(pop_data.P_KOOPWON>0) | (pop_data.P_HUURWON>0)]

    return pop_data

gemeente_pop_data = clean_pop_data(gemeente_pop_data)
wijk_pop_data = clean_pop_data(wijk_pop_data)
buurt_pop_data = clean_pop_data(buurt_pop_data)

# Capitalise column name of variables to merge on (to be the same across data sets)
gemeente_liv_data.rename(columns={"gm_code": "GM_CODE"}, inplace=True)
wijk_liv_data.rename(columns={"wk_code": "WK_CODE"}, inplace=True)
buurt_liv_data.rename(columns={"bu_code": "BU_CODE"}, inplace=True)

# Merge DataFrames
gemeente_data = gemeente_pop_data.merge(gemeente_liv_data, on ="GM_CODE", how = "inner")
wijk_data = wijk_pop_data.merge(wijk_liv_data, on="WK_CODE", how = "inner")
buurt_data = buurt_pop_data.merge(buurt_liv_data, on="BU_CODE", how = "inner")

# Capitalise column name of key variables
gemeente_shape = gemeente_shape.rename(columns={"gm_code": "GM_CODE"}).drop(columns = 'gm_naam')
wijk_shape = wijk_shape.rename(columns={"wk_code": "WK_CODE"}).drop(columns = 'wk_naam')
buurt_shape = buurt_shape.rename(columns={"bu_code": "BU_CODE"}).drop(columns = 'bu_naam')

# Merge the GeoDataFrame and the DataFrame
gemeente_df = gemeente_shape.merge(gemeente_data, on="GM_CODE")
wijk_df = wijk_shape.merge(wijk_data, on="WK_CODE")
buurt_df = buurt_shape.merge(buurt_data, on="BU_CODE")

# CODE YOUR ANSWERS HERE (Use as many cells as you need)
# A
gemeente2provincie_mapping2020 = pd.read_excel(data_folder/'gemeente2provincie_mapping2020.xlsx')

# B
gemeente2provincie_mapping2020.describe()
gemeente2provincie_mapping2020.head()

# C
gemeente_df = gemeente_df.merge(gemeente2provincie_mapping2020,on='GM_CODE')
wijk_df = wijk_df.merge(gemeente2provincie_mapping2020,on='GM_CODE')
buurt_df = buurt_df.merge(gemeente2provincie_mapping2020,on='GM_CODE')
buurt_df['Provincienaam'].head()

0    Groningen
1    Groningen
2    Groningen
3    Groningen
4    Groningen
Name: Provincienaam, dtype: object

# Create histogram and empirical CDF for variable(s) of interest
fig, axes = plt.subplots(1, 2, figsize=(15, 5), sharex=True)
sns.histplot(ax = axes[0],x = wijk_df.lbm)
ecdf_data = sns.ecdfplot(ax = axes[1],x = wijk_df.lbm)
axes[0].set_xlabel("lbm score")
axes[1].set_xlabel("lbm score")
axes[1].grid(True,linewidth = 0.5)
axes[1].minorticks_on()
axes[1].grid(which='minor', linestyle=':', linewidth='0.5', color='black')

# CODE YOUR ANSWERS HERE (Use as many cells as you need)
# A
fig, axes = plt.subplots(1, 2, figsize=(15, 5), sharex=True)
sns.histplot(ax = axes[0],x = wijk_df.soc)
ecdf_data = sns.ecdfplot(ax = axes[1],x = wijk_df.soc)
axes[0].set_xlabel("social cohesion")
axes[1].set_xlabel("social cohesion")
axes[1].grid(True,linewidth = 0.5)
axes[1].minorticks_on()
axes[1].grid(which='minor', linestyle=':', linewidth='0.5', color='black')

# B
# There seems to be a large spread. Furthermore, we see that this variable is 0 centrered

# List top-5 best wijken, based on liveability
wijk_df.nlargest(5,'lbm').head(5)

# CODE YOUR ANSWERS HERE (Use as many cells as you need)
# A
wijk_df.nsmallest(5,'lbm').head(10)

# Plot the liveability in the Netherlands at the wijk-level
fig, ax = plt.subplots(figsize=(60,60))
wijk_df.plot(ax=ax, column = 'lbm', legend = True)
gemeente_df.plot(ax=ax, color  = 'none', edgecolor='black')
_=gemeente_df.apply(lambda x: ax.annotate(text=x['GM_NAAM'], xy=x.geometry.centroid.coords[0], ha='center'), axis=1)
# gemeente_df.apply(lambda x: '' if x['AANT_INW']<50000 else ax.annotate(text=x['GM_NAAM'], xy=x.geometry.centroid.coords[0], ha='center'), axis=1)

# Plot the liveability in Delft at the buurt-level
fig, ax = plt.subplots(figsize=(40,40))
buurt_df.loc[buurt_df.GM_NAAM == 'Delft'].plot(ax=ax, column = 'lbm', legend = True)
gemeente_df.loc[gemeente_df.GM_NAAM == 'Delft'].plot(ax=ax, color  = 'none', edgecolor='black')
buurt_df.loc[buurt_df.GM_NAAM == 'Delft'].apply(lambda x: ax.annotate(text=x['BU_NAAM'], xy=x.geometry.centroid.coords[0], ha='center'), axis=1)
plt.show()

# CODE YOUR ANSWERS HERE (Use as many cells as you need)
# A

fig, ax = plt.subplots(figsize=(40,40))
buurt_df.loc[buurt_df.Provincienaam == 'Zuid-Holland'].plot(ax=ax, column = 'lbm', legend = True)
gemeente_df.loc[gemeente_df.Provincienaam == 'Zuid-Holland'].plot(ax=ax, color  = 'none', edgecolor='black')
gemeente_df.loc[gemeente_df.Provincienaam == 'Zuid-Holland'].apply(lambda x: ax.annotate(text=x['GM_NAAM'], xy=x.geometry.centroid.coords[0], ha='center'), axis=1)
plt.show()

# heatmap of correlations
# Create plot
fig, axes = plt.subplots(figsize=(10, 10))
fig.set_tight_layout(True)

# Compute correlation matrix
corr = wijk_df[['BEV_DICHTH','AANT_INW','AF_SUPERM','AF_RESTAU','P_KOOPWON','P_HUURWON', 'fys', 'onv', 'soc', 'vrz', 'won','lbm']].corr()

# Create upper triangular matrix to mask the upper triangular part of the heatmap
corr_mask = np.triu(np.ones_like(corr, dtype=bool))

# Generate a custom diverging colormap (because it looks better)
corr_cmap = sns.diverging_palette(230, 20, as_cmap=True)

sns.heatmap(corr, mask = corr_mask, cmap=corr_cmap, annot=True,square = True, linewidths=.5, ax = axes)
plt.show()

# Further explore linear associations using scatter plots
# plot
fig, axes = plt.subplots(3, 2, figsize=(20, 10))
fig.set_tight_layout(True)
sns.scatterplot(ax = axes[0,0],x = wijk_df.BEV_DICHTH, y = wijk_df.lbm, alpha = 0.2)
sns.scatterplot(ax = axes[0,1],x = wijk_df.AANT_INW, y = wijk_df.lbm, alpha = 0.2)   
sns.scatterplot(ax = axes[1,0],x = wijk_df.AF_SUPERM, y = wijk_df.lbm, alpha = 0.2)
sns.scatterplot(ax = axes[1,1],x = wijk_df.AF_RESTAU, y = wijk_df.lbm, alpha = 0.2)
sns.scatterplot(ax = axes[2,0],x = wijk_df.P_KOOPWON[wijk_df.P_KOOPWON>0], y = wijk_df.lbm[wijk_df.P_KOOPWON>0], alpha = 0.2)   
sns.scatterplot(ax = axes[2,1],x = wijk_df.P_HUURWON[wijk_df.P_HUURWON>0], y = wijk_df.lbm[wijk_df.P_HUURWON>0], alpha = 0.2)   


axes[0,0].set_xlabel('Population density [#/km^2]')
axes[0,1].set_xlabel('Population size [#]')
axes[1,0].set_xlabel('Distance to supermarket [km]')
axes[1,1].set_xlabel('Distance to restaurant [km]')
axes[2,0].set_xlabel('Percentage owner occupied homes [%]')
axes[2,1].set_xlabel('Percentage rental homes [%]')

Text(0.5, 0, 'Percentage rental homes [%]')

# Box plot
fig, axes = plt.subplots(1, 1, figsize=(15, 10), sharex=True)
GM_select = ['Amsterdam','Rotterdam','Utrecht',"'s-Gravenhage",'Groningen','Tilburg']
df_GM_select = buurt_df[buurt_df['GM_NAAM'].isin(GM_select)]
sns.boxplot(ax = axes, y = df_GM_select.lbm, x = df_GM_select.GM_NAAM, orient="v" )

axes.grid(True,linewidth = 0.5)
plt.show()

# CODE YOUR ANSWERS HERE (Use as many cells as you need)
# A
# Get median values per gemeente
median_lbm_on_wk = wijk_df.groupby('GM_CODE')['lbm'].median()
median_lbm_on_bu = buurt_df.groupby('GM_CODE')['lbm'].median()

fig, axes = plt.subplots(1, 1, figsize=(25, 25))
sns.scatterplot(x = median_lbm_on_wk, y = median_lbm_on_bu)
axes.set_aspect('equal', adjustable='box')
axes.grid(True,linewidth = 0.5)

# Add diagonal line
x_diag = [min(median_lbm_on_bu),max(median_lbm_on_bu)]
y_diag = [min(median_lbm_on_bu),max(median_lbm_on_bu)]
sns.lineplot(x = x_diag, y = x_diag,linestyle = ':',color = 'k')

# Loop through the data points 
for i, gem_name in enumerate(gemeente_df.GM_CODE):
    plt.text(median_lbm_on_wk.iloc[i],median_lbm_on_bu.iloc[i],gemeente_df.GM_NAAM.iloc[i])

# Set the seed number for reproducibility. It governs which data points go to the test set and which go to the training set
rs = 12345

# Create X
X_lin = pd.DataFrame([buurt_df["AANT_INW"].div(100000),buurt_df["BEV_DICHTH"].div(10000), buurt_df["P_KOOPWON"].div(100),buurt_df["fys"],buurt_df["soc"],buurt_df["won"],buurt_df["onv"]]).transpose()


# Create Y
Y = buurt_df.lbm

# Create linear regression object
regr = LinearRegression(fit_intercept = True)

# Split the data in a train and test set
X_train, X_test, Y_train, Y_test = train_test_split(X_lin, Y, test_size=0.25,random_state = rs)

# Fit the model on the training data
regr.fit(X_train,Y_train)

# Evaluate the model generalisation performance on the Train and Test data sets
Y_pred_train = regr.predict(X_train)
mse_train = mean_squared_error(Y_train,Y_pred_train)
R2_train = r2_score(Y_train,Y_pred_train)

Y_pred_test = regr.predict(X_test)
mse_test  = mean_squared_error(Y_test, Y_pred_test)
R2_test = r2_score(Y_test,Y_pred_test)

# Print results
print('Results linear multiple regression model')
print(f'Mean Squared Error Train | Test: \t{mse_train:>7.4f}\t|  {mse_test:>7.4f}')
print(f'R2                 Train | Test: \t{R2_train:>7.4f}\t|  {R2_test:>7.4f}\n')
print('Weights')
print(f'Intercept: \t\t\t {regr.intercept_:>7.4f}')
for n in range(len(regr.coef_)):
    print(f'Weight_{X_lin.keys()[n]:10s} \t\t {regr.coef_[n]:>7.4f}')

Results linear multiple regression model
Mean Squared Error Train | Test: 	 0.0066	|   0.0071
R2                 Train | Test: 	 0.5991	|   0.5820

Weights
Intercept: 			  4.0516
Weight_AANT_INW   		 -0.0626
Weight_BEV_DICHTH 		  0.1351
Weight_P_KOOPWON  		 -0.0259
Weight_fys        		  0.7715
Weight_soc        		  0.3782
Weight_won        		  1.1780
Weight_onv        		  0.6743

# CODE YOUR ANSWERS HERE (Use as many cells as you need)
# A

# CODE YOUR ANSWERS HERE (Use as many cells as you need)
# A
# Your code here
# Use as many cells as you need

# A

# Create data frames
frames1 = pd.DataFrame([buurt_df["AANT_INW"].div(100000),buurt_df["BEV_DICHTH"].div(10000), buurt_df["P_KOOPWON"].div(100),buurt_df["fys"],buurt_df["soc"],buurt_df["won"],buurt_df["onv"]]).transpose()
frames2 = buurt_df[['P_15_24_JR','P_STERFT','P_GESCHEID','STED','WOZ','P_MAROKKO','P_ANT_ARU','P_SURINAM','P_TURKIJE','P_1GEZW','P_MGEZW','OAD','AANT_VROUW']]

X_lin_extended = pd.concat([frames1,frames2],axis=1)
X_lin_extended.replace(-99999999,np.nan, inplace = True)
X_lin_extended.fillna(X_lin_extended.mean(),inplace = True)

# Create Y
Y = buurt_df.lbm

# Create linear regression object
regr = LinearRegression(fit_intercept = True)

# Split the data in a train and test set
X_train, X_test, Y_train, Y_test = train_test_split(X_lin_extended, Y, test_size=0.25,random_state = rs)

# Fit the model on the training data
regr.fit(X_train,Y_train)

# Evaluate the model generalisation performance on the Train and Test data sets
Y_pred_train = regr.predict(X_train)
mse_train = mean_squared_error(Y_train,Y_pred_train)
R2_train = r2_score(Y_train,Y_pred_train)

Y_pred_test = regr.predict(X_test)
mse_test  = mean_squared_error(Y_test, Y_pred_test)
R2_test = r2_score(Y_test,Y_pred_test)

# Print results
print('Results linear multiple regression model')
print(f'Mean Squared Error Train | Test: \t{mse_train:>7.4f}\t|  {mse_test:>7.4f}')
print(f'R2                 Train | Test: \t{R2_train:>7.4f}\t|  {R2_test:>7.4f}\n')
print('Weights')
print(f'Intercept: \t\t\t {regr.intercept_:>7.4f}')
for n in range(len(regr.coef_)):
    print(f'Weight_{X_lin_extended.keys()[n]:10s} \t\t {regr.coef_[n]:>7.4f}')

# B 
# The extended model is probably in region B. We conclude this because we still see little overfitting going on. This suggests the model could further increase in complexity. 
# As a result, we can conclude the first model is probably in Region A. It fits worse that the extended model. This says that it is more biased.

Results linear multiple regression model
Mean Squared Error Train | Test: 	 0.0029	|   0.0028
R2                 Train | Test: 	 0.8251	|   0.8356

Weights
Intercept: 			 10.1865
Weight_AANT_INW   		 -0.9073
Weight_BEV_DICHTH 		 -0.0198
Weight_P_KOOPWON  		  0.0209
Weight_fys        		  0.8517
Weight_soc        		  0.7596
Weight_won        		  1.0501
Weight_onv        		  0.9182
Weight_P_15_24_JR 		 -0.0001
Weight_P_STERFT   		  0.0000
Weight_P_GESCHEID 		 -0.0023
Weight_STED       		 -0.0167
Weight_WOZ        		  0.0000
Weight_P_MAROKKO  		 -0.0003
Weight_P_ANT_ARU  		 -0.0061
Weight_P_SURINAM  		  0.0003
Weight_P_TURKIJE  		 -0.0015
Weight_P_1GEZW    		 -0.0614
Weight_P_MGEZW    		 -0.0607
Weight_OAD        		  0.0001
Weight_AANT_VROUW 		  0.0000

	Gemeentecode	GM_CODE	Gemeentenaam	Provinciecode	ProvinciecodePV	Provincienaam
0	1680	GM1680	Aa en Hunze	22	PV22	Drenthe
1	358	GM0358	Aalsmeer	27	PV27	Noord-Holland
2	197	GM0197	Aalten	25	PV25	Gelderland
3	59	GM0059	Achtkarspelen	21	PV21	Friesland
4	482	GM0482	Alblasserdam	28	PV28	Zuid-Holland

	WK_CODE	gm_naam	geometry	WK_NAAM	GM_CODE	GM_NAAM	IND_WBI	H2O	OAD	STED	...	fys	onv	soc	vrz	won	Gemeentecode	Gemeentenaam	Provinciecode	ProvinciecodePV	Provincienaam
808	WK036302	Amsterdam	POLYGON ((120668.992 487465.601, 120670.593 48...	Grachtengordel-West	GM0363	Amsterdam	1	NEE	10678	1	...	0.003124	-0.057002	-0.057766	0.612074	0.025129	363	Amsterdam	27	PV27	Noord-Holland
828	WK036322	Amsterdam	POLYGON ((120016.875 486171.765, 120130.694 48...	Vondelbuurt	GM0363	Amsterdam	1	NEE	9724	1	...	-0.016662	-0.032454	-0.051559	0.594127	0.027747	363	Amsterdam	27	PV27	Noord-Holland
809	WK036303	Amsterdam	POLYGON ((120757.415 486467.356, 120750.405 48...	Grachtengordel-Zuid	GM0363	Amsterdam	1	NEE	9272	1	...	0.022858	-0.097292	-0.040515	0.607788	0.027745	363	Amsterdam	27	PV27	Noord-Holland
119	WK009600	Vlieland	MULTIPOLYGON (((138217.745 590528.676, 138315....	Wijk 00	GM0096	Vlieland	1	NEE	208	5	...	0.376499	0.122632	0.033750	-0.043568	0.008214	96	Vlieland	21	PV21	Friesland
1291	WK051810	's-Gravenhage	POLYGON ((79575.76 456937.04, 79578.4 456936.4...	Wijk 10 Zorgvliet	GM0518	's-Gravenhage	1	NEE	4163	1	...	0.087135	0.072620	0.014838	0.233574	0.087249	518	's-Gravenhage	28	PV28	Zuid-Holland

	WK_CODE	gm_naam	geometry	WK_NAAM	GM_CODE	GM_NAAM	IND_WBI	H2O	OAD	STED	...	fys	onv	soc	vrz	won	Gemeentecode	Gemeentenaam	Provinciecode	ProvinciecodePV	Provincienaam
69	WK008050	Leeuwarden	POLYGON ((183883.611 580216.812, 183875.967 58...	Heechterp & Schieringen	GM0080	Leeuwarden	1	NEE	2395	2	...	0.032619	-0.241097	-0.078192	-0.009183	-0.064056	80	Leeuwarden	21	PV21	Friesland
1311	WK051830	's-Gravenhage	POLYGON ((79514.87 453802.13, 79503.54 453827....	Wijk 30 Transvaalkwartier	GM0518	's-Gravenhage	1	NEE	7941	1	...	-0.041024	-0.284179	-0.117140	0.234298	-0.146054	518	's-Gravenhage	28	PV28	Zuid-Holland
735	WK034403	Utrecht	POLYGON ((134335.839 458999.112, 134250.809 45...	Wijk 03 Overvecht	GM0344	Utrecht	1	NEE	3276	1	...	-0.030176	-0.246043	-0.087676	0.103762	-0.074552	344	Utrecht	26	PV26	Utrecht
1160	WK047912	Zaanstad	POLYGON ((118678.807 495146.424, 118695.683 49...	Wijk 12 Poelenburg	GM0479	Zaanstad	1	NEE	2176	2	...	-0.008356	-0.185486	-0.073643	0.061053	-0.121695	479	Zaanstad	27	PV27	Noord-Holland
1916	WK085539	Tilburg	POLYGON ((130930.719 399235.829, 131258.299 39...	Wandelbos Noord	GM0855	Tilburg	1	NEE	2018	2	...	-0.019296	-0.158182	-0.055536	-0.023011	-0.065888	855	Tilburg	30	PV30	Noord-Brabant

TPM034A Machine Learning for socio-technical systems¶

`Lab session 01: Discover, explore and visualise data`¶

`Instructions`¶

`Use of AI tools`¶

`Google Colab workspace set-up`¶

`Application: liveability in the Netherlands`

1. Constructing a data set from multiple data sources

A. Loading data

B. Cleaning data¶

C. Combining data¶

Exercise 1: Exploring the data¶

2. Discovering and visualising the data¶

Exercise 3: Face validity of worst neighbourhoods (wijken)¶

Exercise 4: Visualise local neighbourhoods in Zuid-Holland¶

3. Exploring associations¶

Exercise 5: Scatter the median liveability scores of gemeentes based on wijk and buurt level data¶

4. Conducting multiple linear regression analyses¶

Exercise 6: Impact of seed number on the regression results¶

Exercise 7: Generalisation¶

TPM034A Machine Learning for socio-technical systems¶

Lab session 01: Discover, explore and visualise data¶

Instructions¶

Use of AI tools¶

Google Colab workspace set-up¶

Application: liveability in the Netherlands

1. Constructing a data set from multiple data sources

A. Loading data

B. Cleaning data¶

C. Combining data¶

Exercise 1: Exploring the data¶

2. Discovering and visualising the data¶

Exercise 2: Distribution of social cohesion¶

Exercise 3: Face validity of worst neighbourhoods (wijken)¶

Exercise 4: Visualise local neighbourhoods in Zuid-Holland¶

3. Exploring associations¶

Exercise 5: Scatter the median liveability scores of gemeentes based on wijk and buurt level data¶

4. Conducting multiple linear regression analyses¶

Exercise 6: Impact of seed number on the regression results¶

Exercise 7: Generalisation¶

`Lab session 01: Discover, explore and visualise data`¶

`Instructions`¶

`Use of AI tools`¶

`Google Colab workspace set-up`¶

`Application: liveability in the Netherlands`